“Make it look better.” What a seemingly useful quality requirement! But as soon as two people get to it, they have a different definition of “better.” Someone might emphasize its visual appeal, another its speed, and a third its accuracy. All three could be confident they met the same quality requirement, yet produce different results. Acceptance criteria provide an answer to this conundrum by detailing the characteristics a product, service, or end-of-task outcome must possess to qualify for acceptance.
The starting point for acceptance criteria is the desired outcome, not some buzzword such as excellent, reliable, or professional. Consider a common business procedure such as reviewing a finished document before it gets stored. You might ask someone to ensure that a document is correctly filled out. What fields are to be filled out? Does it need to be signed? Does it use the correct date format? Can you do it freehand? More clearly, the acceptance criteria could include: All mandatory fields are filled out, the date format matches the specified format, a valid signature is provided, and no unauthorized modifications are included.
Acceptance criteria must be checkable. It must be clear enough that a person can look at the product and come up with a common conclusion as to whether or not it passes. A document should be neat looks pretty subjective. But a label must be straight, attached, legible and inside the designated area on a package are much clearer criteria. It doesn’t eliminate all subjectivity, but the inspection and evaluation is much more uniform.
Identify one common process step and then list the expected outcome that step should generate. Then cross out all the vague words or phrases. Some common examples that need to be replaced: quickly, properly, correctly, carefully, clean, accurate, appropriate and high quality. You can then replace these with something you can check: observe, count, measure, or audit from a record. An example is quickly becomes within one business day. Data entry accuracy is improved as you copy all required data values from the original record, no fields missing and no formatting changes.
There’s no need to develop the entire process into an exhaustive specification first. Having too many criteria to consider can make a simple check time-consuming and difficult, particularly when you have several criteria to describe the same condition. You should simply look at the first few that separate an acceptable output from a mistake or a deviation. For example, a service request could include the client’s accurate information, a thorough explanation of the incident, an assigned person and a status of completion. You can then add additional criteria as the need develops from an error being repeated.
You’ll want to run the criteria by others. You could just pass it to someone else without giving them a context or definition. Then ask them to observe a work product and decide whether or not it passes. If they pause before they make a decision, if they misinterpret a term, or if they ask you for further clarification before they can make the evaluation, then you need to adjust the wording of the criteria. And you can also apply the same practice for checklists or standard work. It can bring to light any hidden assumptions that might be present before you put it into practice.
Acceptance criteria not only enable the quality process to perform better quality assurance and control checks. They also help improve defect data, problem statements, performance metrics, and corrective actions because the expected result is defined. You don’t have to document that performance is poor, but rather you can say the criteria that weren’t met and the observed condition. If the criteria is strong enough, everyone reviewing the result has the same view of what’s expected, what was observed, and where to focus.